| 1 | //===--- HeaderAnalysis.cpp -------------------------------------*- C++ -*-===// |
| 2 | // |
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | |
| 9 | #include "clang/Tooling/Inclusions/HeaderAnalysis.h" |
| 10 | #include "clang/Basic/SourceLocation.h" |
| 11 | #include "clang/Lex/HeaderSearch.h" |
| 12 | |
| 13 | namespace clang::tooling { |
| 14 | namespace { |
| 15 | |
| 16 | // Is Line an #if or #ifdef directive? |
| 17 | // FIXME: This makes headers with #ifdef LINUX/WINDOWS/MACOS marked as non |
| 18 | // self-contained and is probably not what we want. |
| 19 | bool isIf(llvm::StringRef Line) { |
| 20 | Line = Line.ltrim(); |
| 21 | if (!Line.consume_front(Prefix: "#" )) |
| 22 | return false; |
| 23 | Line = Line.ltrim(); |
| 24 | return Line.starts_with(Prefix: "if" ); |
| 25 | } |
| 26 | |
| 27 | // Is Line an #error directive mentioning includes? |
| 28 | bool isErrorAboutInclude(llvm::StringRef Line) { |
| 29 | Line = Line.ltrim(); |
| 30 | if (!Line.consume_front(Prefix: "#" )) |
| 31 | return false; |
| 32 | Line = Line.ltrim(); |
| 33 | if (!Line.starts_with(Prefix: "error" )) |
| 34 | return false; |
| 35 | return Line.contains_insensitive( |
| 36 | Other: "includ" ); // Matches "include" or "including". |
| 37 | } |
| 38 | |
| 39 | // Heuristically headers that only want to be included via an umbrella. |
| 40 | bool (StringRef Content) { |
| 41 | llvm::StringRef Line; |
| 42 | // Only sniff up to 100 lines or 10KB. |
| 43 | Content = Content.take_front(N: 100 * 100); |
| 44 | for (unsigned I = 0; I < 100 && !Content.empty(); ++I) { |
| 45 | std::tie(args&: Line, args&: Content) = Content.split(Separator: '\n'); |
| 46 | if (isIf(Line) && isErrorAboutInclude(Line: Content.split(Separator: '\n').first)) |
| 47 | return true; |
| 48 | } |
| 49 | return false; |
| 50 | } |
| 51 | |
| 52 | bool isImportLine(llvm::StringRef Line) { |
| 53 | Line = Line.ltrim(); |
| 54 | if (!Line.consume_front(Prefix: "#" )) |
| 55 | return false; |
| 56 | Line = Line.ltrim(); |
| 57 | return Line.starts_with(Prefix: "import" ); |
| 58 | } |
| 59 | |
| 60 | llvm::StringRef getFileContents(FileEntryRef FE, const SourceManager &SM) { |
| 61 | return const_cast<SourceManager &>(SM) |
| 62 | .getMemoryBufferForFileOrNone(File: FE) |
| 63 | .value_or(u: llvm::MemoryBufferRef()) |
| 64 | .getBuffer(); |
| 65 | } |
| 66 | |
| 67 | } // namespace |
| 68 | |
| 69 | bool (FileEntryRef FE, const SourceManager &SM, |
| 70 | const HeaderSearch &) { |
| 71 | if (!HeaderInfo.isFileMultipleIncludeGuarded(File: FE) && |
| 72 | !HeaderInfo.hasFileBeenImported(File: FE) && |
| 73 | // Any header that contains #imports is supposed to be #import'd so no |
| 74 | // need to check for anything but the main-file. |
| 75 | (SM.getFileEntryForID(FID: SM.getMainFileID()) != FE || |
| 76 | !codeContainsImports(Code: getFileContents(FE, SM)))) |
| 77 | return false; |
| 78 | // This pattern indicates that a header can't be used without |
| 79 | // particular preprocessor state, usually set up by another header. |
| 80 | return !isDontIncludeMeHeader(Content: getFileContents(FE, SM)); |
| 81 | } |
| 82 | |
| 83 | bool codeContainsImports(llvm::StringRef Code) { |
| 84 | // Only sniff up to 100 lines or 10KB. |
| 85 | Code = Code.take_front(N: 100 * 100); |
| 86 | llvm::StringRef Line; |
| 87 | for (unsigned I = 0; I < 100 && !Code.empty(); ++I) { |
| 88 | std::tie(args&: Line, args&: Code) = Code.split(Separator: '\n'); |
| 89 | if (isImportLine(Line)) |
| 90 | return true; |
| 91 | } |
| 92 | return false; |
| 93 | } |
| 94 | |
| 95 | std::optional<StringRef> parseIWYUPragma(const char *Text) { |
| 96 | // Skip the comment start, // or /*. |
| 97 | if (Text[0] != '/' || (Text[1] != '/' && Text[1] != '*')) |
| 98 | return std::nullopt; |
| 99 | bool = Text[1] == '*'; |
| 100 | Text += 2; |
| 101 | |
| 102 | // Per spec, direcitves are whitespace- and case-sensitive. |
| 103 | constexpr llvm::StringLiteral IWYUPragma = " IWYU pragma: " ; |
| 104 | if (strncmp(s1: Text, s2: IWYUPragma.data(), n: IWYUPragma.size())) |
| 105 | return std::nullopt; |
| 106 | Text += IWYUPragma.size(); |
| 107 | const char *End = Text; |
| 108 | while (*End != 0 && *End != '\n') |
| 109 | ++End; |
| 110 | StringRef Rest(Text, End - Text); |
| 111 | // Strip off whitespace and comment markers to avoid confusion. This isn't |
| 112 | // fully-compatible with IWYU, which splits into whitespace-delimited tokens. |
| 113 | if (BlockComment) |
| 114 | Rest.consume_back(Suffix: "*/" ); |
| 115 | return Rest.trim(); |
| 116 | } |
| 117 | |
| 118 | } // namespace clang::tooling |
| 119 | |